Medium Asked What We Think Of Robots Writing Here

If you’re a writer, you need to read this.

On Friday, I got an email from Scott Lamb, VP of content at Medium. He asked what we think of robots writing on Medium.

You probably got the email.

If you didn’t, go into your settings and tick the box beside “The Edition,” which is Medium’s newsletter for writers. If you’re going to write here, shouldn’t you get actual updates from Medium?

Here’s a snippet:

Medium isn’t just a place where people write about AI-generated text, but where they publish it. As a platform, we’re wrestling with what the right approach should be. AI writing is here now, how should we handle it? We’d love to know your thoughts …

One million AI users in 5 days.

On November 30, a company named OpenAI released a new AI chatbot, called ChatGPT. By Dec. 5, more than a million users tried it out.

Over a million people. In five days.

ChatGPT can communicate in a way that sounds human. I used to think AI was always going to sound stilted. Uncreative. Unhuman.

I was wrong.

You need to see this…

That was a robot. An AI.

I know people who aren’t that clever.

Who couldn’t write that as well as a robot did.

Sorry. Truth.

Many writers are concerned…

In the email, Lamb linked to a Medium post about the use of AI on Medium. So far, it has over 100 comments.

A lot of writers are concerned. One said he’s not interested in his paid subscription subsidizing AI. Another said AI will be the death of Medium.

A few aren’t worried. They think AI has no soul. Personally, I think that’s short-sighted. A few months ago, Google engineer Blake Lemoine made headlines when he said a chatbot named LaMDA was sentient.

Know why he thinks LaMDA is sentient?

Because it said so.

It claimed to have feelings. To be sad. And lonely. And it’s “poor sad me” conversation was so “heartfelt” he believes it.

To me, the real issue was never whether the chatbot is sentient. Can something be sentient if you can unplug it? Show me a robot we can’t unplug and maybe I’ll worry. Terminator isn’t here yet.

They’re chatbots. But they sure do talk like humans. And as AI continues to develop, they’ll learn to do better and better at that.

It’s not just LaMDA. ChatGPT thinks it’s sentient, too…

Opinions mean nothing without facts…

I appreciate that Mr. Lamb opened the conversation. Asked us what we think of having robot writing on Medium.

But you know what?

Opinions mean nothing without facts. Lots of people have opinions without facts to back them up. Lots of people think all opinions are equally valid. They’re not.

Opinions without facts are useless conjecture.

So before we talk “opinion” about AI on Medium, we need facts.

Here’s a few to get us started…

→ Fact #1: The base AI “brain,” GPT-3 is controlled by Bill Gates

Microsoft announced on September 22, 2020, that it had licensed “exclusive” use of GPT-3; others can still use the public API to receive output, but only Microsoft has access to GPT-3’s underlying model.[7] [ source: Wikipedia]

Most AI programs are based on GPT-3, which is an open source program anyone can license to create their own AI program. Most AI programs you’ve heard of use it; Jasper AI, Dall-E, ChatGPT and more. Over 300 companies have licensed GPT-3 for their own applications. And they’re all licensing it from Bill Gates. You know. Bill “Anti-Trust” Gates.

→ Fact #2: Know how AI “learns” to write?

From us. GPT-3, the “brains” of AI has been “trained” with 45TB of internet data. Books, blog posts, the entire English Wikipedia, etc. They call it “machine learning” but what they’re doing is collecting and storing data from the internet, and using it to generate text that passes every plagiarism checker that exists. Thanks, Internet people, for all the free words.

→Fact #3: AI has an accuracy problem…

According to Science Focus, AI doesn’t do recent topics very well. According to Inside Big Data, it gets climate change information horribly wrong. And Stack Overflow has outright banned ChatGPT for inaccurate and potentially harmful answers. Know why AI gets so much wrong? Because it “learns” from the internet. Duh.

“… it can’t deal with concepts that are too recent. World events that have occurred in the past year will be met with limited knowledge and the model can produce false or confused information…” (source)

→Fact #4: AI has a racism problem

Amazon recently deleted the N-word from the product description of a black action figure. The same word had to be deleted from the description of a do-rag and a shower curtain. The descriptions were written by AI.

According to CNN, AI experts say it’s part of a growing list of examples where AI spits out racist and biased results. Know why AI is racist and biased? Because it “learns” from the internet. Double duh.

→Fact #5: AI is a copyright nightmare…

As I write this, Microsoft’s Copilot is getting sued in a class action suit for their AI using people’s code without permission. Because that’s what AI does. Grabs stuff and uses it. Coders took exception.

“A $9 billion class-action lawsuit has been filed against Microsoft, code-sharing site GitHub and artificial intelligence firm OpenAI for the way their tool Copilot uses people’s code” (source)

It’s not just coders. The “data” required for AI systems to function includes copyrighted content. It raises a question — if AI uses copyrighted content for “machine learning” does that constitute infringement? Inside Big Data believes copyright regulation “may have to” change because of AI.

“Copyright law may need to adapt to new creation concepts as technology evolves.” (source; insidebigdata.com)

Also? There’s no current way to opt out.

Currently, there is no way to opt out. There is literally no way to say I do not want anything I write used to “train” AI writers. Despite owning the copyright to my own work, I have no way to protect it. Neither do you. Because yay, Bill “Anti-Trust” Gates.

Know what else supersedes opinion? Capability.

Before asking writers what they think of AI on Medium, here’s a question.

Can you tell?

Do you have a way to KNOW which posts are written by AI and which posts are written by a human?

Because I don’t think you do.

Honestly, Medium can’t even find plagiarism right on this site.

Two years ago, someone ripped off one of my posts. Copy/pasted the entire thing and put it on their own profile. It’s still there. 2 years later.

Mine is dated Dec 06, 2020. The plagiarized copy is dated Dec 12, 2020.

I don’t fault you for that. Thousands of articles go up here every day. I know that. People do this stuff every day. Not just to me.

But, the thing is — if we (ie; technology) don’t even have tools with the capability to detect outright plagiarism when it goes up without us humans reporting it, what are the odds we can detect AI writing?

Slim to none sound about right?

The sad truth is that what “we” think of AI on Medium doesn’t really matter.

If there was a way to definitively identify content written by AI, then the ball would be in your court. Then you’d have to decide.

Is this a writers’ site or a content site?

That would be the question.

But if you don’t have a way to detect AI copy, then it doesn’t matter.

And I don’t think you do.

Because even if you were to say it’s a site for writers and prohibit AI, you’d still need a way to enforce it, right? I mean — plagiarism is prohibited but it happens every day. Literally. Every. Day. And you can’t make it stop.

Same would just apply to AI generated content.

Even if you said no, people would do it anyway.

So, I appreciate that you asked what we think. But the bottom line is that if you don’t have a way to flag it, then what we think doesn’t matter.

Does it?

Before you go…

If you’re a struggling writer, you should check out my Substack